Development Economics

1
Development Economics

• Risk, Insurance.

2
Risky environment

• Rural households in LDCs are confronted to a
  risky environment
• Climatic risk
• Natural catastrophe
• Illness
• Very variable income over time.
• Spatially correlated risks
• Survival risk for those individual close to the
  subsistence level.

3
Some sources of risk data from Ethiopia
4

• If agents are risk adverse, they get a higher
  utility from a certain income than from an income
  flow with the same expectancy but positive
  variance.
• Hence, they will try to smooth their consumption
  flow.
• Therefore, they need to transfer resources across
  time.

5
Coping with shocks ex-post

• Self-insurance through
• Running down of liquid savings
• Running down of assets stocks (Livestock,
  Jewellery)
• Child labour, school drop out long term
  consequences on human capital accumulation.
• In the long run, might lead to poverty traps.
• In case of aggregate shocks, both asset prices
  and wages are adversely affected.
• Access to credit
• Role of family / migrations

6
(No Transcript)
7
Self-insurance and bullocks Rosenzweig Wolpin
(1993) ICRISAT villages

• In this setting no land sale, but active land
  rental market.
• Largest component of non-land wealth is bullocks
  for bullocks, no rental market but regionally
  integrated market for sales.
• Since there is no rental, there are productivity
  gains to owning bullocks.
• 86 of the HH were involved in a least one
  transaction in bullocks over 10 years, suggesting
  the possibility of distress sales.
• Indeed, because of regional integration, bullock
  prices are immune to village specific production
  shock against which farmers seeks insurance.
• 60 of the sales where made to buyers outside the
  village.
• Results show that the probability of a bullock
  purchase increase with income and probability of
  a sale decreases. Further, farmers holding higher
  stocks of bullocks are less likely to make a
  purchase in the future, suggesting a target
  level.
• RW also suggest that it results in sub-optimal
  holding of bullocks (0.94 instead of 2).

8
Managing risk ex-ante

• Objective consumption smoothing
• Production choices
• Risk diversification
• Share-cropping
• Insurance.

9
Risk sharing

• Possibility of mutual insurance in groups where
  individuals suffer from idiosyncratic shocks.
• A certain amount of consumption smoothing is
  indeed observed (Paxson, Thailand, Townsend,
  India), but no perfect insurance.
• Issue choice of reference group for risk
  sharing. (Munshi Rosenzweig, India cast
  rather than village)

10
Incomplete intertemporal markets

• Informational issue
• Costly, asymmetrical, environment specific.
• Trade-off between risk diversification
  (geographic) and information acquisition.
• Make the implementation of formal insurance
  scheme difficult.
• Gives an advantage to informal mechanisms, in
  particular those relying on family networks
• Migrations (Stark various co-authors).
• Marriages (Rosenzweig Stark ICRISAT)

11
Perfect Insurance model

• At any date, each farmers income is given by
• A is average income, e is a random idiosyncratic
  shock (independant from other farmers shocks),
  and ? is an aggregate shock at the village level.
• If farmers pay the realized value of e into a
  common fund, idiosyncratic shocks can be insured
  against.
• The insured income is then given by
• still fluctuates, but carries less risk
  than Y. The aggregate shock cannot be wiped away
  by mutual insurance within the village.
• Hence, a determinant of mutual insurance is the
  relative importance of idiosyncratic risk to
  aggregate risk.

12
Importance of covariate vs idiosyncratic some
results

• Deaton, Cote dIvoire, finds that common
  components for individual villages explain very
  little of the variation in household income
  changes during 1985-6.
• Townsend, Thailand (1995), finds that there are
  very few common regional components in income
  growth.
• Morduch (2001) studies the ICRISAT villages and
  shows that idiosyncratic risk accounts for 75-96
  percent of the total variance in income.
• Reporting data for villages in Northern Nigeria,
  Udry also finds similar magnitudes of covariant
  risk.

13
Empirical test of perfect insurance

• Controlling for movements in aggregate
  consumption, individual hh income should not
  affect consumption.
• But Hh consumption should follow village average
  consumption
• Regress individual consumption on av. consumption
  and individual income
• cI a ?1 ca ?2 x ?I
• Where the coefficient on the individual income
  shock ?2 is expected to be equal to zero.
• That is, the coefficients on household level
  transitory variations in income should be zero
  and consumption should move one to one with the
  village average (hence ?1 should be equal to 1).

14
Testing for perfect insurance Townsend (1994),
ICRISAT villages again.

• Townsend looks at hh Y average Y this residual
  can be interpreted as idiosyncratic risk (it
  could also include a lot of measurement error).
  It fluctuates independently across hh. hence
  there should be room for mutual insurance.
• Regression of C on Y and average C. He finds
  that indeed coeff of hh income is close to zero,
  and also coeff of sample average C close to one
  (less robust result).
• Regression of distance to the mean in consumption
  on the same in income individual income
  significantly affects individual consumption, but
  the coefficient is small. Estimates elasticity of
  only 14. Further, events such as sickness are
  not significant in determining consumption.
• Smoothing does take place. Impossible to tell
  whether it is from mutual insurance of other
  mechanisms (self-insurance or credit).

15

• Townsend (1994) and Morduch (1995) both find that
  landless household are less able to smooth
  consumption.
• Morduch underlines the hidden cost of smoothing
  household may forego the cultivation of a crop
  with high expected yields because it is more
  risky
• Results on consumption smoothing are also found
  in Thailand (Paxson 1992) and Cote dIvoire
  (Deaton 1994), although full insurance is
  rejected.

16
Limits to insurance information

• Final outcome might not be easily observable so
  that a person might illegitimately claim
  insurance transfers.
• Easier to check if insurer is near, e.g. within
  the village.
• Role of social capital
• Advantage to informal schemes.
• Pb the relevant group is the one within which
  the information flows.

17
Moral hazard.

• In case of full insurance, how is it possible to
  be sure that everyone makes the best effort to
  obtain the high outcome.
• Assume two outcomes H and L
• Probability to obtain H depends on effort with
  effort of cost C, prob is pgt q without effort.
• Without insurance, net expected utility with
  effort is pu(H)(1-p)u(L)-C (1)
• and without effort qu(H)(1-q)u(L)
  (2)
• Assume (1)gt(2).
• If farmers are risk adverse
• u(pH(1-p)L)gtpu(H)(1-p)u(L)

18

• Insuring such level of utility is possible if
  each farmer obtaining H put (1-p)(H-L) into the
  pot, while each farmer with output L receives
  p(H-L). Every one always gets pH(1-p)L for sure.
• Pb how to make sure each one puts effort in.
• Note X consumption of farmer with H and Y the
  consumption of farmer with L.
• An insurance scheme is viable if
• pX(1-p)YpH(1-p)L
• If X and Y are too close, people will not make
  any effort.

19

• How close can they be? Effort will be put in as
  long as
• pu(X)(1-p)u(Y)-Cqu(X)(1-q)u(Y)
• hence the closest they can be is given by
  (p-q)u(X)-u(Y) C
• The second best insurance scheme is a viable
  scheme that respects the above constraint.
• In any solution, XgtY . Hence, the individual
  consumption needs to move with individual income.
  

20

• Here again, group with better information about
  each other are in a better position to provide
  mutual insurance.
• Altruism also facilitates (because it allows
  internalisation of cost to deviation). Hence the
  role of the extended family or kin group.
• Drawback few possibilities of diversification
  within the extended family, except thanks to
  marriage and migration.

21
Enforcement

• High output individuals have incentives to
  deviate. Their gain would be
• Gu(H)-u(M)
• where M is the level of consumption under
  the insurance scheme.
• The loss from deviation is given by
• LNu(M)-pu(H)(1-p)u(L) S
• Where S is measure of utility cost of social
  sanctions and N the number of periods for which
  the insurance arrangement is assumed to hold.

22

• The enforcement constraint requires that LgtG, so
  that
• Nu(M)-pu(H)(1-p)u(L)Su(H)-u(M)
• Role of S
• Role of N perceived instability of an insurance
  scheme is self-fulfilling N is driven down, so
  the enforcement constraint fails and the scheme
  falls apart.
• The net effect of the difference between H and L
  can go either way it both gives incentives to
  deviate but increases the benefits from insurance

23
(No Transcript)
24
Enforcement and imperfect insurance.

• If the enforcement constraint fails, then perfect
  insurance is unattainable.
• The enforcement constraint can be rewritten as
• U(X)Npu(X)(1-p)u(Y)
  u(H)Npu(H)(1-p)u(L)-S
• The RHS term is constant, depends on parameters,
  notably S.
• The LHS term fisrt increases with X and then
  decreases.
• If RHS low enough, perfect insurance is
  sustainable.
• If RHS increases (e.g. because social norms
  weaken), the second best insurance scheme implies
  again a higher X and then a lower Y than under
  perfect insurance. Hence, individual consumption
  has to move with individual income.
• A further increase in RHS will lead to a point
  where no insurance is available.

25

• The blurring between credit and insurance (Udry
  Nigeria) allows to introduce some history
  dependence in the determination of payments a
  contributor at time t will be entitled to a
  bigger share of the returns at time t1.

26
Informal insurance

• Social capital help going around these issues.
• Information, enforcement, repeated relationships,
  social norms, interlinked contracts.
• Examples
• Transfers in Botswana responsive to drought
  (Lucas Stark), to unemployment (livestock
  herding, Cashdan)
• Transfers in India sensitive to shortfalls in
  income (Rosenzweig)
• Transfers from migrant daughter in case of
  illness in the Dominican Republic (De la Brière
  alii)
• Help through loans in India (fisheries, Platteau)

27
Welfare impacts of incomplete insurance

• Rose, India, negative rainfall shocks ? higher
  boy girl mortality in landless households, not
  households with land
• Jacoby Skoufias, Peru, take kids out of school
• Foster, child growth affected during after
  floods in Bangladesh in 1988

28
Welfare impacts of incomplete insurance

• Behrman, India, child health, particularly girls
  negatively affected before major harvest
• Dercon Krishnan, Ethiopia. Intra-hh dimension,
  health shocks result in particularly large losses
  for women 1.6-2.3 percent of body weight